math, pl/math: Reorganise vector routines
Move all AdvSIMD routine sources to math/aarch64/advsimd, and SVE to
math/aarch64/sve. Similarly for headers containing helper routines,
and data tables. Any data files shared between AdvSIMD and SVE remain
in math/aarch64. The same structure is followed under pl/.
Mangling is removed for routine source files, for instance
v_atanf_3u.c now becomes atanf.c. The accuracy tag is not required, as
3.5 ULP is the standard for vector routines, and the v_ prefix is not
required because vector variant is now disambiguated by the
subdirectory. Where routines have non-standard accuracy the tag is
retained.
Mangling for headers containing helper routines etc is straightened
out - we always use v_ prefix rather _advsimd for AdvSIMD, and sv_
rather than _sve for SVE. Though some of these headers, for example
polynomial helpers, could have mangling removed entirely, it is kept
in order to avoid having to make the include paths too fine-grained.
Relative paths (e.g. "..") are removed from includes in order to be
less prescriptive about directory structure.
Also some minor clang-formatting.
199 files changed
